ASX Trading Platform — Review (2026-06-22)

What it is

Buy-and-hold ASX dividend portfolio automation ( $2 K \to$ 200K target). Signal engine with IBKR integration. Control surface at trading.paralla.org. Research-heavy, design-complete, implementation-stalled.

Current state (Phases 1–8 roadmap)

Phase 1 ✅ DONE — ASX universe extraction, fee drag sensitivity, dividend reinvestment study (+485pp DRIP edge over BH), quarterly rebalancing optimal for $2K portfolio
Phase 3 ✅ DONE — fee_gate.py with 24 tests (IBKR AU flat $13.40 round-trip, cost vs expected gain gating, frequency tiers by capital bracket)
Phase 4 ✅ DONE — Risk framework spec + implementation: position_sizer.py (T1/T2/T3 tiers), drawdown_protection.py (no stop-losses, cooling periods at 15%/25%/40% DD), concentration_guard.py (HHI, sector limits), stress_test.py (historical scenarios + Monte Carlo)
Phase 2 ❌ BLOCKED — ibkr-creds blocker. Paper-trading validation stalled. Cannot proceed to live/paper IBKR connection without credentials.
Signal Engine V2 📋 DESIGN ONLY — Architecture doc (signal-engine-v2-design.md) complete: FactorEngine (dividend yield screen + BH), FeeGate integration, DividendLayer for DRIP compounding, MockData pipeline spec. No code written.
Monte Carlo ✅ DONE — Run at $2 K an d$ 10K, 15yr horizon. Results: median final $3, 569/$ 17,843. P(rich) = 0.0% in both cases. Worst DD across all paths: 92.7%.

Gaps / Risks

No end-to-end execution path exists. All modules are isolated scripts/tests. runner.py imports signal_engine but signal engine code doesn’t exist yet (only design).
Phase 8 plan is vague. Original scope mentioned Phases 1–8 but only 1, 3, 4, and partial 5 have deliverables. Phase 2 is dead without creds. Phase 6+ are undocumented.
Paper-trading is theoretical. SimulatedOrderBook in runner.py is standalone; the real IBKR bridge (IBKRExecutionBridge) is untested since ibkr-creds blocked.
No control surface UI. trading.paralla.org has no frontend built yet. Phase 3 was backend-only by design, but no subsequent phase planned for UI.
Monte Carlo results are sobering. 0% probability of reaching $200K target at any starting capital in 15yr horizon with BH+DRIP strategy. This is a fundamental question for the project: does the return target need changing, or is there a different strategy? The project doesn’t address this.
Data quality gap. yfinance covers only ~69% of ASX small-caps. No paid data source procured.
Dependencies not managed in pyproject.toml. Packages (pandas, yfinance, ib_insync) installed via ad-hoc pip calls. Pod restart wipes them. There’s a fix_deps.py script but it’s reactive, not declarative.

Recommended approach

Resolve the blocker first. Get IBKR paper-trading credentials sorted before writing more code. Otherwise Phase 2 stays dead and downstream phases are untestable.
Build signal engine V2 code from the design doc. The architecture is solid — just needs implementation. This is ~5-8 hours of focused work.
Wire runner.py to actually execute a cycle. End-to-end test: data fetch → factor screen → fee gate → risk gate → simulated fill → P&L report. Do this in sim-only mode (no IBKR needed).
Confront the return target. The MC results show BH+DRIP from $2 Kc a n^{'} tpl a u s ib l y hi t$ 200K in 15 years without leverage or higher-risk factors. Decide: accept lower target, add momentum/growth factor, or accept longer horizon.
Define Phases 6–8 explicitly. Don’t leave the roadmap dangling.

Phased plan (recommended)

Phase	Work	Owner	Est.	Depends on
P1-2	Resolve ibkr-creds blocker	apollo/mercury	0.5d	pvs action
P3	Implement signal_engine.py from design	apollo	1d	None
P4	Build mock_data.py (mock pipeline)	apollo	0.5d	P3
P5	Wire runner.py end-to-end (sim mode) + smoke test	apollo	1d	P3, P4
P6	Address return target: revise strategy or accept longer horizon	metis (design)	0.5d	MC results
P7	Declarative deps in pyproject.toml, CI lint	mercury	0.25d	None
P8	Phase 6–8 spec (UI/automation/live go) — design doc only for now	metis	0.5d	All above

Quartz 4

Explorer

ASX Trading Platform — Review